Implementation method for generating transformer substation BIM model by using language based on AI model

Abstract

The invention discloses an implementation method for generating a transformer substation BIM model by using a language based on an AI model, belongs to the technical field of BIM, and fills the blank of conversion from the language to the model: if a main transformer is located on the north side and is connected with a 110kV bus through a circuit breaker, the main transformer is directly converted into a compliant transformer BIM model; aiming at the existing family library of the transformer substation, realizing a searching function of an intelligent agent, accurately calling a searching tool according to parameters input by a user, and obtaining a corresponding equipment model file; and outputting a 3D model file, and strictly ensuring that the file format can be imported into the three-dimensional software (such as obj, fbx, glb and the like) of the Blog super-equal transformer substation for intercommunication.

Description

Technical Field

[0001] The present invention relates to the field of BIM technology, and in particular to a method for implementing a substation BIM model using a language based on an AI model. Background Art

[0002] Over the past two decades, parametric models have been widely used in the power industry. By building parameters, all data and attributes related to power plants, substations, etc. can be associated and adjusted in real time. Parametric models have a more rational and rigorous operating environment, can save manpower and material resources, and facilitate problem detection and adjustment plans. Therefore, they have an inevitable trend in the development of the industry. However, due to the long modeling time, expensive related software training costs, and long training cycles, the development of parametric models has been relatively slow, resulting in low productivity.

[0003] The application of artificial intelligence modeling in the power field is relatively early. Some research institutions and enterprises have carried out research and practice related to AI modeling in this regard, and have released some experimental software. In terms of parametric modeling, many foreign software companies have developed mature software tools, such as Revit. It can realize automated parametric modeling and design optimization. However, parametric modeling and AI modeling have not been linked.

[0004] With the continuous evolution of AI algorithms and the rapid growth of computing needs, the era of artificial intelligence has quietly arrived. However, in the field of power, the application of artificial intelligence is still in its infancy. Currently, the AI ​​model generation software on the market, such as Artefacts.Ai, still generates 3D results that are slightly rough, without model information and normals, and cannot fully meet the requirements of BIM forward design. Most of the popular AI model generation software is mainly used in the entertainment and media industry, and there is still a certain gap in the parametric models required in the power BIM field.

[0005] Based on this technology, a more intuitive and quick model generation and search method was invented, providing an end-to-end solution that automatically generates a substation BIM model that complies with power industry standards through natural language input, significantly improving design efficiency and avoiding human errors. Summary of the invention

[0006] The purpose of the present invention is to provide an implementation method for generating a substation BIM model based on an AI model language in response to the above-mentioned shortcomings, thereby solving the problem that the 3D results generated by the AI ​​model generation software in the prior art are still slightly rough, have no model information and normals, and cannot fully meet the BIM forward design requirements.

[0007] The present invention is achieved through the following scheme: A method for generating a substation BIM model using a language based on an AI model comprises the following steps: Step S1: Demand analysis, professional classification of the BIM models that need to be generated for the substation; Step S2: Parameter determination, dynamically embedding industry specifications through knowledge graphs, and AI automatically avoids more than 90% of human errors; Step S3: Development technology selection, manual evaluation of the accuracy and efficiency of different AI algorithms in processing complex building structures, and whether the modeling tools support interoperability with mainstream substation software; Step S4: System architecture development, developing a natural language parameter extraction tool based on a large model, using natural language processing technology; Step S5: Conducting testing, generating a 3D model based on the input parameters, and applying an optimization algorithm; Step S6: large model training, using the prepared data to train and tune the established artificial intelligence model, feature extraction and conversion of input data, optimization and adjustment of model parameters, and performance evaluation and verification; improve the accuracy of the model through iterative training and tuning; Step S7: Model output and verification, output BIM model files, and strictly ensure that the file format meets the project requirements; use professional verification tools to comprehensively verify the accuracy and completeness of the model from the dimensions of geometric accuracy and topological structure, and use reverse engineering and other technical means to make necessary adjustments and corrections to the problems found until the model meets the substation GIM model delivery standards.

[0008] In step S4, it specifically includes: Step S41, input module: supports text or voice input; Step S42, NLP parsing engine: using a hybrid model that integrates BERT and rule templates to resolve professional terminology ambiguity; Step S43, establishing a knowledge graph library; storing existing family libraries; establishing dynamic rule examples; Step S44, generating modules; topology generation: generating a device connection diagram based on the GraphSAGE model to ensure the correctness of electrical logic; geometry generation: using Stable Diffusion to improve the model, input device parameters and layout constraints, and output a three-dimensional point cloud; Step S45, verification and output module; calling the BIM platform API to generate an IFC file and automatically adding attribute tags.

[0009] In step S43, the specific steps of establishing a knowledge graph library are as follows: 1) Data collection and preprocessing: organize equipment, specifications, and material data; 2) Knowledge graph modeling: define entities, relationships, and dynamic rules; 3) Knowledge graph storage: select a database and import data; 4) Knowledge graph query and verification: use SPARQL or Cypher queries to verify design compliance in real time.

[0010] In step S5, it specifically includes: Step S51, input description; Step S52, generation process; after NLP parsing, the structured data triggers the knowledge graph verification. If the bus height is not specified, the default value is automatically filled in; the diffusion model generates the equipment layout to ensure that the circuit breaker is correctly connected to the transformer and bus segment; Step S53, output results; BIM model: including the three-dimensional geometry and electrical properties of the transformer, circuit breaker, and busbar; automatically generate construction drawings: annotate equipment coordinates, installation instructions, and safety distance detection results.

[0011] In step S6, it specifically includes: Step S61, data collection: collecting existing normative data; Step S62, data enhancement; which includes text data enhancement and image / 3D data enhancement: text data enhancement includes synonym replacement and sentence rewriting; image / 3D data enhancement includes rotation, scaling, and noise injection; Step S63, data storage and management; structured data uses MySQL / PostgreSQL; unstructured data uses MongoDB / MinIO; Step S64, distributed training; data parallelism: split data to multiple GPUs, synchronize gradients; model parallelism: split model layers to different devices; pipeline parallelism: calculate in stages by layer; Step S65, training process optimization.

[0012] In step S65, it specifically includes: reducing video memory usage: mixed precision training, gradient checkpoints, model parallelism; improving throughput: data parallelism, data loading optimization, gradient accumulation; distributed hyperparameter tuning: using Horovod and DeepSpeed ​​for distributed training, combined with Optuna and Ray Tune for hyperparameter search.

[0013] In step S7, it specifically includes: Step S71, BIM model file output Step S72, geometric accuracy verification; Step S73, topology verification; Step S74, problem correction; Step S75, reverse engineering technology derivation; Step S76, AI-assisted correction; Step S77, manual review.

[0014] Step S71, BIM model file output is specifically as follows: 1) Format selection and specification adaptation: Generate BIM model files that conform to the standard format to ensure compatibility with downstream tools; 2) File integrity verification to confirm whether key components exist and whether the model information is complete.

[0015] Step S72, geometric accuracy verification; Specifically: ensure that the generated BIM model meets the design requirements and industry specifications in terms of geometric properties such as size, position, and coordinates; generate a PDF report containing the following: 1) a list of illegal equipment; 2) a heat map of geometric errors; Step S73, topology verification; Specifically: topology verification; ensuring that the equipment connection relationship complies with the design logic and industry standards; 1) electrical connection logic check, including: checking whether the circuit breaker forms a closed loop; verifying whether the rated current of the circuit breaker matches the bus load, etc.; 2) parent-child structure verification; including: checking whether the bus duct belongs to the correct distribution cabinet; verifying whether the child device inherits the parent attribute; Step S74, problem correction; Specific; problem correction; classify problems based on validation results and prioritize critical errors to ensure model compliance; includes: automatic correction and manual intervention.

[0016] Step S75, reverse engineering technology derivation; Specifically, reverse engineering technology is used to derive; by measuring the key dimensions of the generated model, the design parameters are inferred and the original input is optimized; the inferred parameters are input into the generated model and the design is optimized iteratively; Step S76, AI-assisted correction; Specifically, AI assists in correction; training reinforcement learning models to optimize equipment layout and minimize the number of violations; extracting equipment dimensions for generated models; then matching the measured values ​​with equipment library parameters and recommending the closest model; Step S77, manual review; The specific steps are as follows: 1) Visual inspection: check the equipment layout and connection relationship through 3D view; 2) Parameter verification: compare whether the model attribute table is consistent with the design document; 3) Conflict review: use the Clash Detection tool to review structural conflicts.

[0017] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1) Traditional substation BIM modeling relies on manual operation by engineers, and equipment layout, parameters and connection relationships need to be adjusted repeatedly, which takes up to several weeks; accuracy relies on manual verification, and manual secondary correction is required, increasing the risk of errors. In response to this deficiency, the present invention provides a new method for solving substation BIM modeling, filling the gap in language-to-model conversion: such as "the main transformer is located on the north side, connected to the 110kV busbar through a circuit breaker" is directly converted into a compliant transformer BIM model; for the existing family library of the substation, the search function of the intelligent body is realized, and the search tool is accurately called according to the parameters entered by the user to obtain the corresponding equipment model file; the 3D model file is output, and the file format is strictly ensured to be able to be imported into the substation 3D software such as Jinbo Chao for intercommunication (such as .obj, .fbx, .glb, etc.).

[0018] 2) Improved efficiency: The design cycle is shortened from 2-4 weeks of traditional manual modeling to within 10 minutes.

[0019] 3) Reduced error rate: The rule engine can avoid more than 90% of human errors, such as insufficient safety distance.

[0020] 4) Enhanced compatibility: Supports standard formats such as .obj, .fbx, and .glb, and can be directly imported into substation BIM software platforms (such as Revit and Bochao) and power grid simulation systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of the present invention as a whole; Figure 2 is a flow chart of step S4 in the present invention; Figure 3 is a flow chart of step S5 in the present invention; Figure 4 is a flow chart of step S6 in the present invention; Figure 5 It is a flow chart of step S7 in the present invention. DETAILED DESCRIPTION

[0022] All features disclosed in this specification, or steps in all methods or processes disclosed, except mutually exclusive features and / or steps, can be combined in any manner.

[0023] Any feature disclosed in this specification (including any additional claims and abstract), unless otherwise stated, may be replaced by other alternative features that are equivalent or have similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0024] In the description of the present invention, it is necessary to understand that the directions or positional relationships indicated by the terms "up", "down", "left", "right", etc. are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a predetermined direction, be constructed and operated in a predetermined direction, and therefore cannot be understood as a limitation on the present invention.

[0025] In addition, the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first", "second", etc. may explicitly or implicitly include one or more of the features.

[0026] Example 1 like Figure 1~Figure 3 As shown, the present invention provides a technical solution: A method for generating a substation BIM model using a language based on an AI model, comprising at least the following steps: Step S1: Demand analysis, professional classification of the BIM models that need to be generated for the substation, such as electrical primary, electrical secondary, civil engineering, structure, HVAC, water supply and drainage, etc. Step S2: Parameter determination. By dynamically embedding industry standards (such as GB / T 50059) into the knowledge graph, AI can automatically avoid more than 90% of human errors. The automatic avoidance method is: 1) Parse the rules (such as safety distance and equipment compatibility) in standard documents such as GB / T 50059 into structured data; 2) During the model generation process, the knowledge graph rule base is called to verify the design compliance in real time; for example, when generating the main transformer, the safety distance between it and the fence (≥5 meters) is automatically verified; the existing technology usually verifies the model manually, which is prone to errors. AI can automatically avoid more than 90% of human errors.

[0027] Step S3: Development technology selection, manual evaluation of the accuracy and efficiency of different AI algorithms in processing complex building structures, and whether the modeling tools support interoperability with mainstream substation software; Step S4: System architecture development, developing a natural language parameter extraction tool based on a large model, using advanced natural language processing technology, such as the Transformer architecture, to achieve efficient communication and interaction between humans and machines based on natural language; Step S5: Implementing a test, generating a 3D model based on the input parameters, and applying an optimization algorithm to ensure that the model has good program performance while maintaining realism and can run smoothly under different hardware environments; Optimization algorithms can be mesh simplification, texture compression, rendering optimization, or performance testing: Mesh simplification can reduce the number of faces and reduce rendering load; texture compression can reduce the amount of texture data and optimize GPU memory usage; rendering optimization uses LOD, instanced rendering and GPU acceleration technology to improve rendering performance; performance testing can be tested and optimized in different hardware environments to ensure smooth operation; Step S6: large model training, using the prepared data, which is specifically the data in step S1 and the data generated in step S5, to train and tune the established artificial intelligence model, extract and convert the features of the input data, optimize and adjust the model parameters, and evaluate and verify the performance; improve the accuracy of the model through iterative training and tuning; Step S7: Model output and verification, output BIM model files, and strictly ensure that the file format meets the project requirements (such as .obj, .fbx, .glb, etc.); use professional verification tools to comprehensively verify the accuracy and completeness of the model from multiple dimensions such as geometric accuracy and topological structure. For any problems found, use reverse engineering and other technical means to make necessary adjustments and corrections until the model meets the substation GIM model delivery standards.

[0028] Based on the above content, the present invention mainly provides an implementation method for generating a substation BIM model based on an AI model using language, including: developing AI software that can automatically generate a substation BIM model that meets the standards of the power industry through natural language input.

[0029] AI search model, use the search interface of this software to enter keywords to search for existing models in the family library.

[0030] The output of the model format can export the 3D model in a universal format through the software interface, achieving interoperability with commonly used 3D software for substations.

[0031] This solution provides a method of automatically generating a substation BIM model that complies with power industry standards through natural language input; using AI search to input keywords to search for existing models in the family library; and the output model format is interoperable with commonly used 3D software for substations.

[0032] Example 2 like Figure 1~Figure 3 As shown, the present invention provides a technical solution: The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. For steps S1, S2, and S3, they belong to the architectural design and are not within the scope of the invention. The present invention only provides the steps and will not be described in detail. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of the present invention.

[0033] A method for generating a substation BIM model using a language based on an AI model, the method comprising the following steps: S1. Demand analysis; S2. Parameter determination; S3. Selection of development technology; S4. System architecture development; S5. Conduct testing; S6. Large model training; S7. Model output and verification.

[0034] See also Figure 1 , which is a flow chart of the implementation method of generating a substation BIM model based on an AI model language provided by an embodiment of the present invention. In this embodiment, the client is developed using C# language combined with WPF (Windows Presentation Foundation) framework; the backend is developed using Python language combined with FastAPI framework and Blender 3D software. The introduction of this embodiment is based on the above premise; Step S1, demand analysis, professional classification of the BIM models that need to be generated for the substation, such as primary electrical, secondary electrical, civil engineering, structure, HVAC, water supply and drainage, etc.

[0035] Step S2, parameter determination, such as dynamically embedding industry standards through knowledge graphs, such as GB / T 50059, so that AI can automatically avoid more than 90% of human errors.

[0036] Step S3, development technology selection, evaluates the accuracy and efficiency of different AI algorithms in dealing with complex building structures, and whether the modeling tools support interoperability with mainstream substation software.

[0037] Step S4, system architecture development, develop a natural language parameter extraction tool based on a large model, and use advanced natural language processing technologies, such as the Transformer architecture, to achieve efficient communication and interaction between humans and machines based on natural language.

[0038] Specifically include: Step S41, input module. Support text or voice input, such as "110kV indoor station, main transformer capacity 50MVA, distance from west wall ≥8 meters"; Step S42, NLP parsing engine: Use a hybrid model that integrates BERT and rule templates to resolve ambiguity in professional terms, such as "busbar" refers to copper bus or cable; output structured data with probability confidence, and request clarification from the user when the confidence is low; In this solution, "NLP" is the abbreviation of Natural Language Processing, which is called natural language processing in Chinese. It is a branch of artificial intelligence (AI) and computer science that focuses on enabling computers to understand, interpret, and generate human language. In this solution, "BERT" (Bidirectional Encoder Representation from Transformers) is a pre-training model proposed by Google AI Research in October 2018.

[0039] Step S43, establish a knowledge graph library; store existing family libraries, such as: equipment library (transformer, circuit breaker, etc.), specification library (national standard, industry standard), material library (size, electrical parameters); establish dynamic rule examples, such as: IF equipment type = circuit breaker THEN the connection object must contain bus AND transformer.

[0040] This solution establishes dynamic rules, which can transform the organic structure of equipment or specifications, such as parameters such as equipment grade and size, into a database to form specifications, so that the generated model complies with the specifications.

[0041] The specific steps are as follows: 1) Data collection and preprocessing: Organize equipment, specifications, and material data; 2) Knowledge graph modeling: Define entities, relationships, and dynamic rules; 3) Knowledge graph storage: Select a database and import data; 4) Knowledge graph query and verification: Use SPARQL or Cypher queries to verify design compliance in real time; Step S44, generating a module; Topology generation: Generate device connection diagram based on GraphSAGE model to ensure electrical logic correctness; Geometry generation: Use Stable Diffusion to improve the model, input device parameters and layout constraints, and output 3D point cloud; Step S45, verification and output module; call the BIM platform API to generate an IFC file, automatically add attribute tags, such as: equipment ID, operation and maintenance information; output violation reports and correction suggestions, such as: "The distance between the main transformer and the wall is insufficient, it is recommended to move it 2.1 meters eastward"; Step S5, conducting a test, generating a 3D model based on the input parameters, and applying optimization algorithms, such as mesh simplification, texture compression, and other technologies, to ensure that the model has good program performance while maintaining realism and can run smoothly under different hardware environments; Specifically include: Step S51, input description; such as: "35kV substation, including two dry-type transformers (capacity 10MVA), located on the east side of the plant, with a spacing of ≥3 meters, connected to a single busbar segment through a vacuum circuit breaker"; Step S52, generation process; after NLP parsing, the structured data triggers the knowledge graph verification. If the bus height is not specified, the default value of 4.5 meters is automatically filled in; the diffusion model generates the equipment layout, and the GNN ensures that the circuit breaker correctly connects the transformer and the bus segment; in this solution, GNN, that is, graph neural network, is a deep learning technology for processing graph data; Step S53, output results; BIM model (IFC format): including the three-dimensional geometry and electrical properties of the transformer, circuit breaker, and busbar; automatically generate construction drawings (PDF): annotate equipment coordinates, installation instructions, and safety distance detection results; Step S6, large model training; use the prepared data, which is specifically the data in step S1 and the data generated in step S5, to train and tune the established artificial intelligence model; improve the accuracy of the model through iterative training and tuning; Specifically Step S61, data collection: collecting existing standard data, such as power industry standard data and equipment parameter library; Step S62, data enhancement; text data: synonym replacement, sentence rewriting (applicable to natural language input); image / 3D data: rotation, scaling, noise injection (applicable to BIM model generation).

[0042] Step S63, data storage and management; structured data uses MySQL / PostgreSQL (equipment parameters, design specifications); unstructured data uses MongoDB / MinIO (text description, BIM files).

[0043] Step S64, distributed training; data parallelism: split data to multiple GPUs, synchronize gradients; model parallelism: split model layers to different devices; pipeline parallelism: calculate in stages by layer; Step S65, optimizing the training process, such as reducing video memory usage, improving throughput, and using a distributed hyperparameter tuning framework. Specific: Reduce video memory usage: mixed precision training, gradient checkpoints, model parallelism. Improve throughput: data parallelism, data loading optimization, gradient accumulation. Distributed hyperparameter tuning: use Horovod and DeepSpeed ​​for distributed training, and combine Optuna and Ray Tune for hyperparameter search.

[0044] Step S7, model output and verification; output BIM model files, strictly ensure that the file format meets the project requirements (such as .obj, .fbx, .glb, etc.); use professional verification tools to comprehensively verify the accuracy and completeness of the model from multiple dimensions such as geometric accuracy and topological structure; use reverse engineering and other technical means to make necessary adjustments and corrections to the problems found until the model meets the substation GIM model delivery standards.

[0045] Specific: Step S71, BIM model file output; including: 1) format selection and specification adaptation: generating a BIM model file that complies with the standard format (such as .obj, .fbx, .glb) to ensure compatibility with downstream tools; 2) file integrity verification, such as: whether key components exist and whether the model information is complete; Step S72, geometric accuracy verification; such as: size and position, object coordinates.

[0046] Specifically: Verify geometric accuracy; ensure that the generated BIM model meets the design requirements and industry specifications in terms of geometric properties such as size, position, and coordinates; generate a PDF report containing the following: 1) a list of illegal equipment (such as coordinate offset, size discrepancy); 2) a heat map of geometric errors (visual display of problem areas).

[0047] Step S73: topology verification, such as electrical connection logic, parent-child structure.

[0048] Specifically: topology structure verification; ensuring that the equipment connection relationship (such as electrical topology, hierarchical structure) complies with the design logic and industry standards; 1) Electrical connection logic check, including: checking whether the circuit breaker forms a closed loop (violating the single bus segmentation design); verifying whether the rated current of the circuit breaker matches the bus load, etc.; 2) Parent-child structure verification; including: checking whether the bus duct belongs to the correct distribution cabinet; verifying whether the child device inherits the parent attribute (such as voltage level).

[0049] Step S74: problem correction, such as problem classification and priority sorting.

[0050] Specific; problem correction; classify problems based on validation results and prioritize critical errors to ensure model compliance; includes: automatic correction and manual intervention.

[0051] Step S75, reverse engineering technology derivation; for example, by measuring the key dimensions of the generated model, the design parameters are deduced.

[0052] Specifically, reverse engineering technology is used for derivation; by measuring the key dimensions of the generated model, the design parameters are inferred and the original input is optimized; the inferred parameters are input into the generated model and the design is optimized iteratively.

[0053] Step S76: AI-assisted correction. Train the reinforcement learning model to optimize the layout and minimize the number of violations.

[0054] Specifically, AI assists in correction; trains reinforcement learning models to optimize equipment layout and minimize the number of violations; for example: extracts the equipment dimensions of the generated model (such as the length, width and height of the transformer); then matches the measured values ​​with the equipment library parameters and recommends the closest model.

[0055] Step S77: Manual review. BIM engineer review.

[0056] Specifically, manual review; BIM engineer review; the steps are as follows: 1) Visual inspection: check the equipment layout and connection relationship through 3D view; 2) Parameter verification: compare the model attribute table with the design document for consistency; 3) Conflict review: use the Clash Detection tool to review structural conflicts.

[0057] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for generating a substation BIM model using a language based on an AI model, characterized in that: The following steps are involved: Step S1: Demand analysis, professional classification of the BIM models that need to be generated for the substation; Step S2: Parameter determination, dynamically embedding industry specifications through knowledge graphs, and AI automatically avoids more than 90% of human errors; Step S3: Development technology selection, manual evaluation of the accuracy and efficiency of different AI algorithms in processing complex building structures, and whether the modeling tools support interoperability with mainstream substation software; Step S4: System architecture development, developing a natural language parameter extraction tool based on a large model, using natural language processing technology; Step S5: Conducting testing, generating a 3D model based on the input parameters, and applying an optimization algorithm; Step S6: large model training, using the prepared data to train and tune the established artificial intelligence model, feature extraction and conversion of input data, optimization and adjustment of model parameters, and performance evaluation and verification; improve the accuracy of the model through iterative training and tuning; Step S7: Model output and verification, output BIM model files, and strictly ensure that the file format meets the project requirements; use professional verification tools to comprehensively verify the accuracy and completeness of the model from the dimensions of geometric accuracy and topological structure, and use reverse engineering technology to make necessary adjustments and corrections to the problems found until the model meets the substation GIM model delivery standards.

2. According to claim 1, a method for generating a substation BIM model using a language based on an AI model is characterized by: In step S4, it specifically includes: Step S41, input module: supports text or voice input; Step S42, NLP parsing engine: using a hybrid model that integrates BERT and rule templates to resolve professional terminology ambiguity; Step S43, establishing a knowledge graph library; storing existing family libraries; establishing dynamic rule examples; Step S44, generating modules; topology generation: generating a device connection diagram based on the GraphSAGE model to ensure the correctness of electrical logic; geometry generation: using Stable Diffusion to improve the model, input device parameters and layout constraints, and output a three-dimensional point cloud; Step S45, verification and output module; calling the BIM platform API to generate an IFC file and automatically adding attribute tags.

3. According to claim 2, a method for generating a substation BIM model using a language based on an AI model is characterized by: In step S43, the specific steps of establishing a knowledge graph library are as follows: 1) Data collection and preprocessing: organize equipment, specifications, and material data; 2) Knowledge graph modeling: define entities, relationships, and dynamic rules; 3) Knowledge graph storage: select a database and import data; 4) Knowledge graph query and verification: use SPARQL or Cypher queries to verify design compliance in real time.

4. A method for generating a substation BIM model using a language based on an AI model according to any one of claims 1 to 3, characterized in that: In step S5, it specifically includes: Step S51, input description; Step S52, generation process; after NLP parsing, the structured data triggers the knowledge graph verification. If the bus height is not specified, the default value is automatically filled in; the diffusion model generates the equipment layout to ensure that the circuit breaker is correctly connected to the transformer and bus segment; Step S53, output results; BIM model: including the three-dimensional geometry and electrical properties of the transformer, circuit breaker, and busbar; automatically generate construction drawings: annotate equipment coordinates, installation instructions, and safety distance detection results.

5. A method for generating a substation BIM model using a language based on an AI model according to any one of claims 1 to 3, characterized in that: In step S6, it specifically includes: Step S61, data collection: collecting existing normative data; Step S62, data enhancement; which includes text data enhancement and image / 3D data enhancement: text data enhancement includes synonym replacement and sentence rewriting; image / 3D data enhancement includes rotation, scaling, and noise injection; Step S63, data storage and management; structured data uses MySQL / PostgreSQL; unstructured data uses MongoDB / MinIO; Step S64, distributed training; data parallelism: split data to multiple GPUs, synchronize gradients; model parallelism: split model layers to different devices; pipeline parallelism: calculate in stages by layer; Step S65, training process optimization.

6. According to claim 5, a method for generating a substation BIM model using a language based on an AI model is characterized by: In step S65, specifically including: reducing video memory usage: mixed precision training, gradient checkpoints, model parallelism; Improve throughput: data parallelism, data loading optimization, and gradient accumulation; Distributed hyperparameter tuning: Use Horovod and DeepSpeed ​​for distributed training, and combine Optuna and Ray Tune for hyperparameter search.

7. A method for generating a substation BIM model using a language based on an AI model according to any one of claims 1 to 3, characterized in that: In step S7, it specifically includes: Step S71, BIM model file output; Step S72, geometric accuracy verification; Step S73, topology verification; Step S74, problem correction; Step S75, reverse engineering technology derivation; Step S76, AI-assisted correction; Step S77, manual review.

8. The method for generating a substation BIM model using a language based on an AI model according to claim 7 is characterized in that: Step S71, BIM model file output is specifically as follows: 1) Format selection and specification adaptation: Generate BIM model files that conform to the standard format to ensure compatibility with downstream tools; 2) File integrity verification to confirm whether key components exist and whether the model information is complete.

9. The method for generating a substation BIM model using a language based on an AI model according to claim 7, characterized in that: Step S72, geometric accuracy verification; Specifically: ensure that the geometric properties of the generated BIM model in terms of size, position, and coordinates meet the design requirements and industry specifications; generate a PDF report containing the following: 1) a list of illegal equipment; 2) a heat map of geometric errors; Step S73, topology verification; Specifically: topology verification; ensuring that the equipment connection relationship complies with the design logic and industry standards; 1) electrical connection logic check, including: checking whether the circuit breaker forms a closed loop; verifying whether the rated current of the circuit breaker matches the bus load; 2) parent-child structure verification; including: checking whether the bus duct belongs to the correct distribution cabinet; verifying whether the child device inherits the parent attribute; Step S74, problem correction; Specific; problem correction; classify problems based on validation results and prioritize critical errors to ensure model compliance; includes: automatic correction and manual intervention.

10. The method for generating a substation BIM model using a language based on an AI model according to claim 7, characterized in that: Step S75, reverse engineering technology derivation; Specifically, reverse engineering technology is used to derive; by measuring the key dimensions of the generated model, the design parameters are inferred and the original input is optimized; the inferred parameters are input into the generated model and the design is optimized iteratively; Step S76, AI-assisted correction; Specifically, AI-assisted correction; Train a reinforcement learning model to optimize device layout and minimize the number of violations; extract device dimensions for the generated model; then match the measurements with the device library parameters and recommend the closest model; Step S77, manual review; The specific steps are as follows: 1) Visual inspection: check the equipment layout and connection relationship through 3D view; 2) Parameter verification: compare whether the model attribute table is consistent with the design document; 3) Conflict review: use the Clash Detection tool to review structural conflicts.

Images